Research on Adaptive Parameters Determination in DBSCAN Algorithm

摘要

Two parameters Eps and MinPts are required to be inputted manually in DBSCAN algorithm, and this tedious intervention leads to the situation that the clustering precision depends largely on user's entry. This paper proposed a new method to determine the two parameters, which can avoid the manual intervention, and even realize the clustering automatically. Experimental results show that the method can determine the two parameters more reasonably. Furthermore, it can get clustering results more accurately.